08/07/2021 Erik Kusch
1
Erik Kusch (erik.kusch@bio.au.dk), PhD
Student
Department of Biology
Section for Ecoinformatics & Biodiversity
Center for Biodiversity Dynamics in a
Changing World (BIOCHANGE)
Aarhus University
Erik Kusch
1
STATISTICAL
EDUCATION FOR
BIOLOGISTS
Why are biologists rarely statistically literate?
09/09/2021 SalGo-Team Statistical Education
08/07/2021 Erik Kusch
2
09/09/2021
THE PROBLEMS IN STATISTICAL
EDUCATION FOR BIOLOGISTS
Bees in my Bonnet
SalGo-Team Statistical Education
08/07/2021 Erik Kusch
3Expectations & Mentality
SalGo-Team Statistical Education09/09/2021
Disregard for statistics among students:
“It is no fun”
“R does not like me”
“I will not need statistics”
“I do not want to do research with a lot of statistics”
Disconnected from the reality of academia
Disregard among colleagues:
“We have always done our stats in Excel. I do not see any reason to change that.”
“Correlation is the same as causation as long as the hypothesis tested is simple.”
Mentality ranging from calcified to incorrect
Underappreciation for Biostatistics
08/07/2021 Erik Kusch
4
Biology has a lot of subdisciplines :
Botany
Zoology
Microbiology
Biotechnology
Genetics
Etc.
Statistics are relevant for all of these!
B.Sc. Studies focus heavily on “old-school” biology
Lack of Focus & Outdated Curriculum
SalGo-Team Statistical Education09/09/2021
Teaching of irrelevant knowledge
08/07/2021 Erik Kusch
5
B.Sc. Studies make many subdisciplines of biology obligatory courses
Enormous amounts of time wasted
Statistical Education often treated as a by-product of other courses
No one takes time out of their schedule to explain a different topic
Lack of Time & Effort
SalGo-Team Statistical Education09/09/2021
Subpar statistical education
08/07/2021 Erik Kusch
6
Little time devoted to teaching Biostats
Rushed Teaching
Large volume of knowledge needs conveying
Shallow Knowledge
Rushed Teaching & Shallow Knowledge
SalGo-Team Statistical Education09/09/2021
Incorrect use of Biostatistics
08/07/2021 Erik Kusch
7
R and Statistics are usually taught at the same time
Total overload for students
Simultaneous teaching of R and Statistics reduces time
spent on either
Even shallower knowledge
Learning R & Statistics
SalGo-Team Statistical Education09/09/2021
Resentment for biostatistics
08/07/2021 Erik Kusch
8
09/09/2021
TIME FOR CHANGE
How do we address these issues?
SalGo-Team Statistical Education
08/07/2021 Erik Kusch
9
STOP advertising university research and careers to potential students exclusively with cute animals and
aesthetically pleasing plants
Sets wrong expectations
Makes students resent their studies when they have to deal with desk work
START communicating the work in our department(s) accurately to potential students
Expectations match reality
Students that do sign on will moan much less about biostatistics and subsequently perform better
Advertising Biological Studies
SalGo-Team Statistical Education09/09/2021
08/07/2021 Erik Kusch
10
Make more “biological” courses elective
Make more biostatistics courses obligatory
Proposed minimum time spent on statistics:
B.Sc.: ~ ½ year (assuming a 3-year program)
M.Sc.: ~ ¼ year (assuming a 2-year program)
PhD: However long is necessary for the PhD project
Time-Budgeting
SalGo-Team Statistical Education09/09/2021
08/07/2021 Erik Kusch
11
Deepen course contents:
Statistical Practices and Guidelines
Data Handling
Model Assumptions & Violations
Math!
Course Contents
SalGo-Team Statistical Education09/09/2021
08/07/2021 Erik Kusch
12
Separate Teaching R from Biostatistics!
Teaching R
SalGo-Team Statistical Education09/09/2021
08/07/2021 Erik Kusch
13
09/09/2021
THE PERFECT1COURSE
CURRICULUM
1The following are super personal opinions and wishes
SalGo-Team Statistical Education
08/07/2021 Erik Kusch
14
Course Credits / Time Contents
Leve
l
Obligator
y
R
Language of
Computation
5ECTS / 140h
Data Handling & Data Mining with R
Data Management Practices
B.Sc.
🗸
Introduction to
Biostatistics
10ECTS / 280h
Practices in Statistics
Descriptive Statistics vs. Inferential Statistics
Nominal/Correlation/Ordinal and Metric Tests
Simple Parametric Tests
🗸
Linear Regression
10ECTS / 280h
Linear Regression & Assumptions
Fitting Methods (Least-Squares vs. ML)
Variable Selection
Model Selection/Comparison
🗸
Classifications
5ECTS / 140h
Classification (KNN, LDA, QDA, Hierarchies, RF,
Networks, Logistic Regression)
Gradient Analyses (Direct, Indirect)
🗸
Data Visualisation
5ECTS / 140h
Plot Types & Their Applications
Plotting with R
B.Sc. Level
SalGo-Team Statistical Education09/09/2021
08/07/2021 Erik Kusch
15 M.Sc. & PhD Level
09/09/2021
Course Credits / Time Contents
Leve
l
Obligator
y
Beyond Linearity
10ECTS / 280h
Polynomial Regression, Step/Basis Functions
Regression/Smoothing Splines
Mixed Effect Models, GLS/GLM and GAM
M.Sc
.
🗸
Spatio
-Temporal
Analyses
10ECTS / 280h
Geospatial Models
Spatial Autocorrelation
SDMs
Interpolation
Time-Series Analyses
Temporal Autocorrelation
Lag-Effects, and Granger Causality
Statistical &
Machine Learning
5ECTS / 140h
SVMs, Neural Networks
Resampling Methods
Model Regularisation
High-Dimensionality
🗸
Bayesian Statistics
10ECTS / 280h
Bayesian Theory & Application PhD
Data Integration
5ECTS / 140h
Data Repositories & Model Integration
08/07/2021 Erik Kusch
16
SalGo-Team Statistical Education09/09/2021